Search tips
Search criteria

Results 1-25 (81)

Clipboard (0)

Select a Filter Below

more »
Year of Publication
more »
1.  Using methods from the data mining and machine learning literature for disease classification and prediction: A case study examining classification of heart failure sub-types 
Journal of clinical epidemiology  2013;66(4):398-407.
Physicians classify patients into those with or without a specific disease. Furthermore, there is often interest in classifying patients according to disease etiology or subtype. Classification trees are frequently used to classify patients according to the presence or absence of a disease. However, classification trees can suffer from limited accuracy. In the data-mining and machine learning literature, alternate classification schemes have been developed. These include bootstrap aggregation (bagging), boosting, random forests, and support vector machines.
Study design and Setting
We compared the performance of these classification methods with those of conventional classification trees to classify patients with heart failure according to the following sub-types: heart failure with preserved ejection fraction (HFPEF) vs. heart failure with reduced ejection fraction (HFREF). We also compared the ability of these methods to predict the probability of the presence of HFPEF with that of conventional logistic regression.
We found that modern, flexible tree-based methods from the data mining literature offer substantial improvement in prediction and classification of heart failure sub-type compared to conventional classification and regression trees. However, conventional logistic regression had superior performance for predicting the probability of the presence of HFPEF compared to the methods proposed in the data mining literature.
The use of tree-based methods offers superior performance over conventional classification and regression trees for predicting and classifying heart failure subtypes in a population-based sample of patients from Ontario. However, these methods do not offer substantial improvements over logistic regression for predicting the presence of HFPEF.
PMCID: PMC4322906  PMID: 23384592
Boosting; classification trees; Bagging; random forests; classification; regression trees; support vector machines; regression methods; statistical methods; prediction; heart failure
2.  Impact of Modality Choice on Rates of Hospitalization in Patients Eligible for Both Peritoneal Dialysis and Hemodialysis 
♦ Background: Hospitalization rates are a relevant consideration when choosing or recommending a dialysis modality. Previous comparisons of peritoneal dialysis (PD) and hemodialysis (HD) have not been restricted to individuals who were eligible for both therapies.
♦ Methods: We conducted a multicenter prospective cohort study of people 18 years of age and older who were eligible for both PD and HD, and who started outpatient dialysis between 2007 and 2010 in four Canadian dialysis programs. Zero-inflated negative binomial models, adjusted for baseline patient characteristics, were used to examine the association between modality choice and rates of hospitalization.
♦ Results: The study enrolled 314 patients. A trend in the HD group toward higher rates of hospitalization, observed in the primary analysis, became significant when modality was treated as a time-varying exposure or when the population was restricted to elective outpatient starts in patients with at least 4 months of pre-dialysis care. Cardiovascular disease, infectious complications, and elective surgery were the most common reasons for hospital admission; only 23% of hospital stays were directly related to complications of dialysis or kidney disease.
♦ Conclusions: Efforts to promote PD utilization are unlikely to result in increased rates of hospitalization, and efforts to reduce hospital admissions should focus on potentially avoidable causes of cardiovascular disease and infectious complications.
PMCID: PMC3923691  PMID: 24525596
Hemodialysis; hospitalization
3.  Is pathology necessary to predict mortality among men with prostate-cancer? 
Statistical models developed using administrative databases are powerful and inexpensive tools for predicting survival. Conversely, data abstraction from chart review is time-consuming and costly. Our aim was to determine the incremental value of pathological data obtained from chart abstraction in addition to information acquired from administrative databases in predicting all-cause and prostate cancer (PC)-specific mortality.
We identified a cohort of men with diabetes and PC utilizing population-based data from Ontario. We used the c-statistic and net-reclassification improvement (NRI) to compare two Cox- proportional hazard models to predict all-cause and PC-specific mortality. The first model consisted of covariates from administrative databases: age, co-morbidity, year of cohort entry, socioeconomic status and rural residence. The second model included Gleason grade and cancer volume in addition to all aforementioned variables.
The cohort consisted of 4001 patients. The accuracy of the admin-data only model (c-statistic) to predict 5-year all-cause mortality was 0.7 (95% CI 0.69-0.71). For the extended model (including pathology information) it was 0.74 (95% CI 0.73-0.75). This corresponded to a change in category of predicted probability of survival among 14.8% in the NRI analysis.
The accuracy of the admin-data model to predict 5-year PC specific mortality was 0.76 (95% CI 0.74-0.78). The accuracy of the extended model was 0.85 (95% CI 0.83-0.87). Corresponding to a 28% change in the NRI analysis.
Pathology chart abstraction, improved the accuracy in predicting all-cause and PC-specific mortality. The benefit is smaller for all-cause mortality, and larger for PC-specific mortality.
PMCID: PMC4275978
Prostate cancer; Survival; Prediction models; Population-based study
4.  Using fractional polynomials to model the effect of cumulative duration of exposure on outcomes: applications to cohort and nested case-control designs 
Determining the nature of the relationship between cumulative duration of exposure to an agent and the hazard of an adverse outcome is an important issue in environmental and occupational epidemiology, public health and clinical medicine. The Cox proportional hazards regression model can incorporate time-dependent covariates. An important class of continuous time-dependent covariates is that denoting cumulative duration of exposure.
We used fractional polynomial methods to describe the association between cumulative duration of exposure and adverse outcomes. We applied these methods in a cohort study to examine the relationship between cumulative duration of use of the antiarrhythmic drug amiodarone and the risk of thyroid dysfunction. We also used these methods with a conditional logistic regression model in a nested case-control study to examine the relationship between cumulative duration of use of bisphosphonate medication and the risk of atypical femur fracture.
Using a cohort design and a Cox proportional hazards model, we found a non-linear relationship between cumulative duration of use of the antiarrhythmic drug amiodarone and the risk of thyroid dysfunction. The risk initially increased rapidly with increasing cumulative use. However, as cumulative duration of use increased, the rate of increase in risk attenuated and eventually levelled off. Using a nested case-control design and a conditional logistic regression model, we found evidence of a linear relationship between duration of use of bisphosphonate medication and risk of atypical femur fractures.
Fractional polynomials allow one to model the relationship between cumulative duration of medication use and adverse outcomes.
PMCID: PMC4230473  PMID: 24664670
Cox proportional hazards regression model; fractional polynomials; time-dependent covariates; survival analysis; pharmacoepidemiology; cohort study; nested case-control study
5.  Association between appropriateness of coronary revascularization and quality of life in patients with stable ischemic heart disease 
The relationship between appropriateness score, treatment strategy and quality of life (QOL) among patients with stable ischemic heart disease (SIHD) is not known. In this prospective cohort study, we evaluated changes in generic and cardiac-specific quality of life in patients with documented SIHD, comparing patients with revascularization versus those with medical therapy alone, stratified by their appropriateness scores.
Consecutive patients with SIHD undergoing elective coronary angiogram from November 1st 2008 to December 1st 2009 completed the Seattle Angina Questionnaire (SAQ) and EQ-5D at the time of procedure and at 1 year. The appropriateness for coronary revascularization was determined at the time of coronary angiography.
Our final cohort consisted of 425 patients, 69.4% of whom underwent revascularization. In the overall cohort, 272 (64.0%) had appropriate indications for revascularization, while 57 (13.4%) had inappropriate indications and 96 (22.6%) had uncertain indications. On average, patients improved in most QOL domains, regardless of treatment strategy and appropriateness score. In patients with appropriate indications, revascularized patients had greater improvements in both generic (0.073; 95% CI 0.003-0.144; p-value 0.04) and disease-specific indices, including angina stability (14.6; 95% CI 0.85-28.3; p-value 0.04), physical limitation (9.3; 95% CI 0.71-17.8; p-value 0.03) and disease perception (12.7; 95% CI4.3-21.1; p-value 0.003) compared to medically treated patients. However, patients with uncertain and inappropriate indications also had improvements in physical limitation and disease perception with revascularization compared to medical therapy.
Patients who had appropriate revascularization derived the greatest improvement in QOL compared with medical therapy.
Electronic supplementary material
The online version of this article (doi:10.1186/1471-2261-14-137) contains supplementary material, which is available to authorized users.
PMCID: PMC4195906  PMID: 25280534
Quality of life; Angina; Appropriateness
6.  Association Between Metformin Therapy and Mortality After Breast Cancer 
Diabetes Care  2013;36(10):3018-3026.
Metformin has been associated with a reduction in breast cancer risk and may improve survival after cancer through direct and indirect tumor-suppressing mechanisms. The purpose of this study was to evaluate the effect of metformin therapy on survival in women with breast cancer using methods that accounted for the duration of treatment with glucose-lowering therapies.
This population-based study, using Ontario health care databases, recruited women aged 66 years or older diagnosed with diabetes and breast cancer between 1 April 1997 and 31 March 2008. Using Cox regression analyses, we explored the association between cumulative duration of past metformin use and all-cause and breast cancer–specific mortality. We modeled cumulative duration of past metformin use as a time-varying exposure.
Of 2,361 breast cancer patients identified, mean (± SD) age at cancer diagnosis was 77.4 ± 6.3 years, and mean follow-up was 4.5 ± 3.0 years. There were 1,101 deaths(46.6%), among which 386 (16.3%) were breast cancer–specific deaths. No significant association was found between cumulative duration of past metformin use and all-cause mortality (adjusted hazard ratio 0.97 [95% CI 0.92–1.02]) or breast cancer–specific mortality (0.91 [0.81–1.03]) per additional year of cumulative use.
Our findings failed to show an association between improved survival and increased cumulative metformin duration in older breast cancer patients who had recent-onset diabetes. Further research is needed to clarify this association, accounting for effects of cancer stage and BMI in younger populations or those with differing stages of diabetes as well as in nondiabetic populations.
PMCID: PMC3781496  PMID: 23633525
7.  Cardiovascular Complications and Mortality After Diabetes Diagnosis for South Asian and Chinese Patients 
Diabetes Care  2013;36(9):2670-2676.
Many non-European ethnic groups have an increased risk for diabetes; however, the published literature demonstrates considerable uncertainty about the rates of diabetes complications among minority populations. The objective of this study was to determine the risks of cardiovascular complications and of mortality after diabetes diagnosis for South Asian and Chinese patients, compared with European patients.
A population-based cohort study identified all 491,243 adults with newly diagnosed diabetes in Ontario, Canada, between April 2002 and March 2009. Subjects were followed until March 2011 for the first occurrence of any cardiovascular complication of diabetes (coronary artery disease, stroke, or lower-extremity amputation) and for all-cause mortality. Median follow-up was 4.7 years.
The crude incidence of cardiovascular complications after diabetes diagnosis was 17.9 per 1,000 patient-years among European patients, 12.0 among South Asian patients, and 7.7 among Chinese patients. After adjusting for baseline characteristics, the cause-specific hazard ratios (HRs) for cardiovascular complications relative to European patients were 0.95 (95% CI 0.90–1.00; P = 0.056) and 0.50 (0.46–0.53; P < 0.001) for South Asian and Chinese patients, respectively. Mortality was lower for both minority groups (adjusted HR for South Asian patients 0.56 [95% CI 0.52–0.60]; P < 0.001; for Chinese patients 0.58 [0.55–0.62]; P < 0.001).
Chinese patients were at substantially lower risk than European patients for cardiovascular complications after diabetes diagnosis, whereas South Asian patients were at comparable risk. Mortality after diabetes diagnosis was markedly lower for both minority populations.
PMCID: PMC3747942  PMID: 23637350
8.  A Tutorial and Case Study in Propensity Score Analysis: An Application to Estimating the Effect of In-Hospital Smoking Cessation Counseling on Mortality 
Multivariate Behavioral Research  2011;46(1):119-151.
Propensity score methods allow investigators to estimate causal treatment effects using observational or nonrandomized data. In this article we provide a practical illustration of the appropriate steps in conducting propensity score analyses. For illustrative purposes, we use a sample of current smokers who were discharged alive after being hospitalized with a diagnosis of acute myocardial infarction. The exposure of interest was receipt of smoking cessation counseling prior to hospital discharge and the outcome was mortality with 3 years of hospital discharge. We illustrate the following concepts: first, how to specify the propensity score model; second, how to match treated and untreated participants on the propensity score; third, how to compare the similarity of baseline characteristics between treated and untreated participants after stratifying on the propensity score, in a sample matched on the propensity score, or in a sample weighted by the inverse probability of treatment; fourth, how to estimate the effect of treatment on outcomes when using propensity score matching, stratification on the propensity score, inverse probability of treatment weighting using the propensity score, or covariate adjustment using the propensity score. Finally, we compare the results of the propensity score analyses with those obtained using conventional regression adjustment.
PMCID: PMC3266945  PMID: 22287812
9.  Printed educational messages aimed at family practitioners fail to increase retinal screening among their patients with diabetes: a pragmatic cluster randomized controlled trial [ISRCTN72772651] 
Evidence of the effectiveness of printed educational messages in narrowing the gap between guideline recommendations and practice is contradictory. Failure to screen for retinopathy exposes primary care patients with diabetes to risk of eye complications. Screening is initiated by referral from family practitioners but adherence to guidelines is suboptimal. We aimed to evaluate the ability of printed educational messages aimed at family doctors to increase retinal screening of primary care patients with diabetes.
Design: Pragmatic 2×3 factorial cluster trial randomized by physician practice, involving 5,048 general practitioners (with 179,833 patients with diabetes). Setting: Ontario family practitioners. Interventions: Reminders (that retinal screening helps prevent diabetes-related vision loss and is covered by provincial health insurance for patients with diabetes) with prompts to encourage screening were mailed to each physician in conjunction with a widely-read professional newsletter. Alternative printed materials formats were an ‘outsert’ (short, directive message stapled to the outside of the newsletter), and/or a two-page, evidence-based article (‘insert’) and a pre-printed sticky note reminder for patients. Main outcome measure: A successful outcome was an eye examination (which includes retinal screening) provided to a patient with diabetes, not screened in the previous 12 months, within 90 days after visiting a family practitioner. Analysis accounted for clustering of doctors within practice groups.
No intervention effect was detected (eye exam rates were 31.6% for patients of control physicians, 31.3% for the insert, 32.8% for the outsert, 32.3% for those who received both, and 31.2% for those who received both plus the patient reminder with the largest 95% confidence interval around any effect extending from −1.3% to 1.1%).
This large trial conclusively failed to demonstrate any impact of printed educational messages on screening uptake. Despite their low cost, printed educational messages should not be routinely used in attempting to close evidence-practice gaps relating to diabetic retinopathy screening.
Trial registration
Electronic supplementary material
The online version of this article (doi:10.1186/1748-5908-9-87) contains supplementary material, which is available to authorized users.
PMCID: PMC4261896  PMID: 25098587
10.  Using Ensemble-Based Methods for Directly Estimating Causal Effects: An Investigation of Tree-Based G-Computation 
Multivariate behavioral research  2012;47(1):115-135.
Researchers are increasingly using observational or nonrandomized data to estimate causal treatment effects. Essential to the production of high-quality evidence is the ability to reduce or minimize the confounding that frequently occurs in observational studies. When using the potential outcome framework to define causal treatment effects, one requires the potential outcome under each possible treatment. However, only the outcome under the actual treatment received is observed, whereas the potential outcomes under the other treatments are considered missing data. Some authors have proposed that parametric regression models be used to estimate potential outcomes. In this study, we examined the use of ensemble-based methods (bagged regression trees, random forests, and boosted regression trees) to directly estimate average treatment effects by imputing potential outcomes. We used an extensive series of Monte Carlo simulations to estimate bias, variance, and mean squared error of treatment effects estimated using different ensemble methods. For comparative purposes, we compared the performance of these methods with inverse probability of treatment weighting using the propensity score when logistic regression or ensemble methods were used to estimate the propensity score. Using boosted regression trees of depth 3 or 4 to impute potential outcomes tended to result in estimates with bias equivalent to that of the best performing methods. Using an empirical case study, we compared inferences on the effect of in-hospital smoking cessation counseling on subsequent mortality in patients hospitalized with an acute myocardial infarction.
PMCID: PMC3293511  PMID: 22419832 CAMSID: cams2143
12.  A Tutorial and Case Study in Propensity Score Analysis: An Application to Estimating the Effect of In-Hospital Smoking Cessation Counseling on Mortality 
Multivariate behavioral research  2011;46(1):119-151.
Propensity score methods allow investigators to estimate causal treatment effects using observational or nonrandomized data. In this article we provide a practical illustration of the appropriate steps in conducting propensity score analyses. For illustrative purposes, we use a sample of current smokers who were discharged alive after being hospitalized with a diagnosis of acute myocardial infarction. The exposure of interest was receipt of smoking cessation counseling prior to hospital discharge and the outcome was mortality with 3 years of hospital discharge. We illustrate the following concepts: first, how to specify the propensity score model; second, how to match treated and untreated participants on the propensity score; third, how to compare the similarity of baseline characteristics between treated and untreated participants after stratifying on the propensity score, in a sample matched on the propensity score, or in a sample weighted by the inverse probability of treatment; fourth, how to estimate the effect of treatment on outcomes when using propensity score matching, stratification on the propensity score, inverse probability of treatment weighting using the propensity score, or covariate adjustment using the propensity score. Finally, we compare the results of the propensity score analyses with those obtained using conventional regression adjustment.
PMCID: PMC3266945  PMID: 22287812 CAMSID: cams1834
13.  An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies 
Multivariate Behavioral Research  2011;46(3):399-424.
The propensity score is the probability of treatment assignment conditional on observed baseline characteristics. The propensity score allows one to design and analyze an observational (nonrandomized) study so that it mimics some of the particular characteristics of a randomized controlled trial. In particular, the propensity score is a balancing score: conditional on the propensity score, the distribution of observed baseline covariates will be similar between treated and untreated subjects. I describe 4 different propensity score methods: matching on the propensity score, stratification on the propensity score, inverse probability of treatment weighting using the propensity score, and covariate adjustment using the propensity score. I describe balance diagnostics for examining whether the propensity score model has been adequately specified. Furthermore, I discuss differences between regression-based methods and propensity score-based methods for the analysis of observational data. I describe different causal average treatment effects and their relationship with propensity score analyses.
PMCID: PMC3144483  PMID: 21818162
14.  Boosted classification trees result in minor to modest improvement in the accuracy in classifying cardiovascular outcomes compared to conventional classification trees 
Purpose: Classification trees are increasingly being used to classifying patients according to the presence or absence of a disease or health outcome. A limitation of classification trees is their limited predictive accuracy. In the data-mining and machine learning literature, boosting has been developed to improve classification. Boosting with classification trees iteratively grows classification trees in a sequence of reweighted datasets. In a given iteration, subjects that were misclassified in the previous iteration are weighted more highly than subjects that were correctly classified. Classifications from each of the classification trees in the sequence are combined through a weighted majority vote to produce a final classification. The authors' objective was to examine whether boosting improved the accuracy of classification trees for predicting outcomes in cardiovascular patients. Methods: We examined the utility of boosting classification trees for classifying 30-day mortality outcomes in patients hospitalized with either acute myocardial infarction or congestive heart failure. Results: Improvements in the misclassification rate using boosted classification trees were at best minor compared to when conventional classification trees were used. Minor to modest improvements to sensitivity were observed, with only a negligible reduction in specificity. For predicting cardiovascular mortality, boosted classification trees had high specificity, but low sensitivity. Conclusions: Gains in predictive accuracy for predicting cardiovascular outcomes were less impressive than gains in performance observed in the data mining literature.
PMCID: PMC3253503  PMID: 22254181
Boosting; classification trees; predictive model; classification; recursive partitioning; congestive heart failure; acute myocardial infarction; outcomes research
15.  Radiographic monitoring of incidental abdominal aortic aneurysms: a retrospective population-based cohort study 
Open Medicine  2011;5(2):e77-79.
An abdominal aortic aneurysm (AAA) that is identified when the abdomen is imaged for some other reason is known as an incidental AAA. No population-based studies have assessed the management of incidental AAAs. The objective of this study was to measure the completeness of radiographic monitoring of incidental AAAs by means of a population-based analysis.
We linked a cohort of patients with incidental AAA (defined as a previously unidentified aortic enlargement exceeding 30 mm in diameter found in an imaging study performed for another reason) to various population-based databases. We followed the patients to elective repair or rupture of the aneurysm, death or 31 Mar. 2009. We used evidence-based monitoring guidelines to calculate the proportion of observation time during which each incidental AAA was incompletely monitored. We used negative binomial regression to determine the association of patient-related factors with this outcome.
For the period between January 1996 and September 2008, we identified 191 patients with incidental AAA (mean diameter 37.6 mm, 95% confidence interval [CI] 36.6–38.6 mm; median follow-up 4.4 [range 0.6–12.7] years). Fifty-six of these patients (29.3%) had no radiographic monitoring of the aneurysm. Overall, patients spent one-fifth of their time with incomplete monitoring of the AAA (median 19.4%, interquartile range 0.3%–44.0%). Factors independently associated with incomplete monitoring included older age (relative rate [change in proportion of time with incomplete monitoring] [RR] 1.27, 95% CI 1.10–1.47, per decade), larger size (RR 1.65, 95% CI 1.38–2.01, per 10-mm increase) and detection of the aneurysm while the patient was in hospital or the emergency department (RR 1.34, 95% CI 1.00–1.79). Comorbidities were not associated with monitoring.
Radiographic monitoring of incidental AAAs was incomplete, and almost one-third of patients underwent no monitoring at all. Incomplete monitoring did not appear to be related to patients’ comorbidity.
PMCID: PMC3147999  PMID: 21915236
16.  Effects of impairment in activities of daily living on predicting mortality following hip fracture surgery in studies using administrative healthcare databases 
BMC Geriatrics  2014;14:9.
Impairment in activities of daily living (ADL) is an important predictor of outcomes although many administrative databases lack information on ADL function. We evaluated the impact of ADL function on predicting postoperative mortality among older adults with hip fractures in Ontario, Canada.
Sociodemographic and medical correlates of ADL impairment were first identified in a population of older adults with hip fractures who had ADL information available prior to hip fracture. A logistic regression model was developed to predict 360-day postoperative mortality and the predictive ability of this model were compared when ADL impairment was included or omitted from the model.
The study sample (N = 1,329) had a mean age of 85.2 years, were 72.8% female and the majority resided in long-term care (78.5%). Overall, 36.4% of individuals died within 360 days of surgery. After controlling for age, sex, medical comorbidity and medical conditions correlated with ADL impairment, addition of ADL measures improved the logistic regression model for predicting 360 day mortality (AIC = 1706.9 vs. 1695.0; c -statistic = 0.65 vs 0.67; difference in - 2 log likelihood ratios: χ2 = 16.9, p = 0.002).
Direct measures of ADL impairment provides additional prognostic information on mortality for older adults with hip fractures even after controlling for medical comorbidity. Observational studies using administrative databases without measures of ADLs may be potentially prone to confounding and bias and case-mix adjustment for hip fracture outcomes should include ADL measures where these are available.
PMCID: PMC3922692  PMID: 24472282
Hip fracture; Predictive models; Mortality; Outcomes; Activities of daily living; Confounding; Risk adjustment
17.  Statistical Criteria for Selecting the Optimal Number of Untreated Subjects Matched to Each Treated Subject When Using Many-to-One Matching on the Propensity Score 
American Journal of Epidemiology  2010;172(9):1092-1097.
Propensity-score matching is increasingly being used to estimate the effects of treatments using observational data. In many-to-one (M:1) matching on the propensity score, M untreated subjects are matched to each treated subject using the propensity score. The authors used Monte Carlo simulations to examine the effect of the choice of M on the statistical performance of matched estimators. They considered matching 1–5 untreated subjects to each treated subject using both nearest-neighbor matching and caliper matching in 96 different scenarios. Increasing the number of untreated subjects matched to each treated subject tended to increase the bias in the estimated treatment effect; conversely, increasing the number of untreated subjects matched to each treated subject decreased the sampling variability of the estimated treatment effect. Using nearest-neighbor matching, the mean squared error of the estimated treatment effect was minimized in 67.7% of the scenarios when 1:1 matching was used. Using nearest-neighbor matching or caliper matching, the mean squared error was minimized in approximately 84% of the scenarios when, at most, 2 untreated subjects were matched to each treated subject. The authors recommend that, in most settings, researchers match either 1 or 2 untreated subjects to each treated subject when using propensity-score matching.
PMCID: PMC2962254  PMID: 20802241
bias (epidemiology); matching; Monte Carlo method; observational study; propensity score
18.  Speak Fast, Use Jargon, and Don’t Repeat Yourself: A Randomized Trial Assessing the Effectiveness of Online Videos to Supplement Emergency Department Discharge Instructions 
PLoS ONE  2013;8(11):e77057.
Emergency department discharge instructions are variably understood by patients, and in the setting of emergency department crowding, innovations are needed to counteract shortened interaction times with the physician. We evaluated the effect of viewing an online video of diagnosis-specific discharge instructions on patient comprehension and recall of instructions.
In this prospective, single-center, randomized controlled trial conducted between November 2011 and January 2012, we randomized emergency department patients who were discharged with one of 38 diagnoses to either view (after they left the emergency department) a vetted online video of diagnosis-specific discharge instructions, or to usual care. Patients were subsequently contacted by telephone and asked three standardized questions about their discharge instructions; one point was awarded for each correct answer. Using an intention-to-treat analysis, differences between groups were assessed using univariate testing, and with logistic regression that accounted for clustering on managing physician. A secondary outcome measure was patient satisfaction with the videos, on a 10-point scale.
Among 133 patients enrolled, mean age was 46.1 (s.d.D. 21.5) and 55% were female. Patients in the video group had 19% higher mean scores (2.5, s.d. 0.7) than patients in the control group (2.1, s.d. 0.8) (p=0.002). After adjustment for patient age, sex, first language, triage acuity score, and clustering, the odds of achieving a fully correct score (3 out of 3) were 3.5 (95% CI, 1.7 to 7.2) times higher in the video group, compared to the control group. Among those who viewed the videos, median rating of the videos was 10 (IQR 8 to 10).
In this single-center trial, patients who viewed an online video of their discharge instructions scored higher on their understanding of key concepts around their diagnosis and subsequent care. Those who viewed the videos found them to be a helpful addition to standard care.
Trial Registration NCT01361932
PMCID: PMC3823877  PMID: 24244272
19.  Impact of provider volume on operative mortality after radical cystectomy in a publicly funded healthcare system 
We assess the effect of cystectomy provider volume on postoperative mortality in a publicly funded healthcare system. Hospital and surgeon (provider) volume have been shown to be associated with clinically important outcomes for many types of surgery. Volume-outcome studies in patients undergoing radical cystectomy for bladder cancer have primarily originated from privately funded healthcare systems.
We identified patients undergoing cystectomy in Ontario, Canada, between 1992 and 2004 using administrative databases. The effect of provider volume on postoperative mortality was assessed with multilevel (hierarchical or random effects) logistic regression models, adjusted for patient characteristics. Separate models were fit to examine the effect of surgeon volume and the effect of hospital volume.
Of the 3296 cystectomy patients identified, 126 (3.8%) experienced a postoperative death. Neither hospital volume (odds ratio [per 1 unit increase in volume] 0.98, 95% confidence interval [CI] 0.95–1.00; p = 0.074) nor surgeon volume (odds ratio 0.96, 95% CI 0.90–1.02; p = 0.143) were statistically significantly associated with postoperative cystectomy mortality.
In Ontario’s publicly funded healthcare system, provider volume was not significantly associated with postoperative mortality.
PMCID: PMC3876442  PMID: 24381661
20.  The Average Lifespan of Patients Discharged from Hospital with Heart Failure 
Journal of General Internal Medicine  2012;27(9):1171-1179.
There are no life-tables quantifying the average life-spans of post-hospitalized heart failure populations across various strata of risk.
To quantify the life-expectancies (i.e., average life-spans) of heart failure patients at the time of hospital discharge according to age, gender, predictive 30-day mortality heart failure risk index, and comorbidity burden.
Population-based retrospective cohort study.
Ontario, Canada.
7,865 heart failure patients discharged from Ontario hospitals between 1999 and 2000.
Data were obtained from the Enhanced Feedback for Effective Cardiac Treatment EFFECT provincial quality improvement initiative. All patients were linked to administrative data, and tracked longitudinally until March 31, 2010. Detailed clinical variables were obtained from medical chart abstraction, and death data were obtained from vital statistics. Average life-spans were calculated using Cox Proportion Hazards models in conjunction with the Declining Exponential Approximation of Life Expectancy (D.E.A.L.E) method to extrapolate life-expectancy, adjusting for age, gender, predicted 30-day mortality, left ventricular function and comorbidity, and was reported according to key prognostic risk-strata.
The average life-span of the cohort was 5.5 years (STD +/− 10.0) ranging from 19.5 years for low-risk women of less than 50 years old to 2.9 years for high-risk octogenarian males. Average life-spans were lower by 0.13 years among patients with impaired as compared with preserved left ventricular function, and by approximately one year among patients with three or more as compared with no concomitant comorbidities. In total, 17.4 % and 27 % of patients had died within 6 months and 1 year respectively, despite having predicted life-spans exceeding one-year.
Data regarding changes in patient clinical status over time were unavailable.
The development of risk-adjusted life-tables for heart failure populations is feasible and mirrored those with advanced malignant diseases. Average life span varied widely across clinical risk strata, and may be less accurate among those at or near their end of life.
PMCID: PMC3515002  PMID: 22549300
life-expectancy; heart failure; risk; survival; comorbidity
21.  Discriminating clinical features of heart failure with preserved vs. reduced ejection fraction in the community 
European Heart Journal  2012;33(14):1734-1741.
Heart failure (HF) is a major public health burden worldwide. Of patients presenting with HF, 30–55% have a preserved ejection fraction (HFPEF) rather than a reduced ejection fraction (HFREF). Our objective was to examine discriminating clinical features in new-onset HFPEF vs. HFREF.
Methods and results
Of 712 participants in the Framingham Heart Study (FHS) hospitalized for new-onset HF between 1981 and 2008 (median age 81 years, 53% female), 46% had HFPEF (EF >45%) and 54% had HFREF (EF ≤45%). In multivariable logistic regression, coronary heart disease (CHD), higher heart rate, higher potassium, left bundle branch block, and ischaemic electrocardiographic changes increased the odds of HFREF; female sex and atrial fibrillation increased the odds of HFPEF. In aggregate, these clinical features predicted HF subtype with good discrimination (c-statistic 0.78). Predictors were examined in the Enhanced Feedback for Effective Cardiac Treatment (EFFECT) study. Of 4436 HF patients (median age 75 years, 47% female), 32% had HFPEF and 68% had HFREF. Distinguishing clinical features were consistent between FHS and EFFECT, with comparable discrimination in EFFECT (c-statistic 0.75). In exploratory analyses examining the traits of the intermediate EF group (EF 35–55%), CHD predisposed to a decrease in EF, whereas other clinical traits showed an overlapping spectrum between HFPEF and HFREF.
Multiple clinical characteristics at the time of initial HF presentation differed in participants with HFPEF vs. HFREF. While CHD was clearly associated with a lower EF, overlapping characteristics were observed in the middle of the left ventricular EF range spectrum.
PMCID: PMC3530391  PMID: 22507977
Heart failure; Epidemiology; Risk factors; Ejection fraction
22.  Socioeconomic Status, Functional Recovery, and Long-Term Mortality among Patients Surviving Acute Myocardial Infarction 
PLoS ONE  2013;8(6):e65130.
To examine the relationship between socio-economic status (SES), functional recovery and long-term mortality following acute myocardial infarction (AMI).
The extent to which SES mortality disparities are explained by differences in functional recovery following AMI is unclear.
We prospectively examined 1368 patients who survived at least one-year following an index AMI between 1999 and 2003 in Ontario, Canada. Each patient was linked to administrative data and followed over 9.6 years to track mortality. All patients underwent medical chart abstraction and telephone interviews following AMI to identify individual-level SES, clinical factors, processes of care (i.e., use of, and adherence, to evidence-based medications, physician visits, invasive cardiac procedures, referrals to cardiac rehabilitation), as well as changes in psychosocial stressors, quality of life, and self-reported functional capacity.
As compared with their lower SES counterparts, higher SES patients experienced greater functional recovery (1.80 ml/kg/min average increase in peak V02, P<0.001) after adjusting for all baseline clinical factors. Post-AMI functional recovery was the strongest modifiable predictor of long-term mortality (Adjusted HR for each ml/kg/min increase in functional capacity: 0.91; 95% CI: 0.87–0.94, P<0.001) irrespective of SES (P = 0.51 for interaction between SES, functional recovery, and mortality). SES-mortality associations were attenuated by 27% after adjustments for functional recovery, rendering the residual SES-mortality association no longer statistically significant (Adjusted HR: 0.84; 95% CI:0.70–1.00, P = 0.05). The effects of functional recovery on SES-mortality associations were not explained by access inequities to physician specialists or cardiac rehabilitation.
Functional recovery may play an important role in explaining SES-mortality gradients following AMI.
PMCID: PMC3670842  PMID: 23755180
23.  Absence of Disparities in the Quality of Primary Diabetes Care for South Asians and Chinese in an Urban Canadian Setting 
Diabetes Care  2012;35(4):794-796.
To examine whether quality of diabetes care is equitable for South Asian and Chinese patients in an urban Canadian setting.
Process and intermediate measures of quality of care were compared between 246 South Asians, 170 Chinese, and 431 patients from the general population with type 2 diabetes selected from 45 family physicians’ practices.
A total of 61% of Chinese achieved A1C ≤7.0% versus 45% of South Asians and 49% of the general population (P < 0.05). They were also more likely to achieve LDL cholesterol ≤2.0 mmol/L, while South Asians were more likely to achieve blood pressure ≤130/80. There was only one significant process of care deficiency: fewer foot examinations among South Asians (34 vs. 49% for the general population, P < 0.01).
Quality of diabetes care in a Canadian urban setting was equitable, with ethnic minorities somewhat more likely to achieve recommended targets than the general population.
PMCID: PMC3308276  PMID: 22323411
24.  The Timing of Drug Funding Announcements Relative to Elections: A Case Study Involving Dementia Medications 
PLoS ONE  2013;8(2):e56921.
Following initial regulatory approval of prescription drugs, many factors may influence insurers and health systems when they decide whether to add these drugs to their formularies. The role of political pressures on drug funding announcements has received relatively little attention, and elections represent an especially powerful form of political pressure. We examined the temporal relationship between decisions to add one class of drugs to publicly funded formularies in Canada's ten provinces and elections in these jurisdictions.
Dates of provincial formulary listings for cholinesterase inhibitors, which are drugs used to treat Alzheimer's disease and related dementias, were compared to the dates of provincial elections. Medical journal articles, media reports, and proceedings from provincial legislatures were reviewed to assemble information on the chronology of events. We tested whether there was a statistically significant increase in the probability of drug funding announcements within the 60-day intervals preceding provincial elections.
Decisions to fund the cholinesterase inhibitors were made over a nine-year span from 1999 to 2007 in the ten provinces. In four of ten provinces, the drugs were added to formularies in a time period closely preceding a provincial election (P = 0.032); funding announcements in these provinces were made between 2 and 47 days prior to elections. Statements made in provincial legislatures highlight the key role of political pressures in these funding announcements.
Impending elections appeared to affect the timing of drug funding announcements in this case study. Despite an established structure for evidence-based decision-making, drug funding remains a complex process open to influence from many sources. Awareness of such influences is critical to maintain effective drug policy and public health decision-making.
PMCID: PMC3584056  PMID: 23460820
25.  Cardiovascular disease after Escherichia coli O157:H7 gastroenteritis 
Escherichia coli O157:H7 is one cause of acute bacterial gastroenteritis, which can be devastating in outbreak situations. We studied the risk of cardiovascular disease following such an outbreak in Walkerton, Ontario, in May 2000.
In this community-based cohort study, we linked data from the Walkerton Health Study (2002–2008) to Ontario’s large healthcare databases. We included 4 groups of adults: 3 groups of Walkerton participants (153 with severe gastroenteritis, 414 with mild gastroenteritis, 331 with no gastroenteritis) and a group of 11 263 residents from the surrounding communities that were unaffected by the outbreak. The primary outcome was a composite of death or first major cardiovascular event (admission to hospital for acute myocardial infarction, stroke or congestive heart failure, or evidence of associated procedures). The secondary outcome was first major cardiovascular event censored for death. Adults were followed for an average of 7.4 years.
During the study period, 1174 adults (9.7%) died or experienced a major cardiovascular event. Compared with residents of the surrounding communities, the risk of death or cardiovascular event was not elevated among Walkerton participants with severe or mild gastroenteritis (hazard ratio [HR] for severe gastroenteritis 0.74, 95% confidence interval [CI] 0.38–1.43, mild gastroenteritis HR 0.64, 95% CI 0.42–0.98). Compared with Walkerton participants who had no gastroenteritis, risk of death or cardiovascular event was not elevated among participants with severe or mild gastroenteritis.
There was no increase in the risk of cardiovascular disease in the decade following acute infection during a major E. coli O157:H7 outbreak.
PMCID: PMC3537814  PMID: 23166291

Results 1-25 (81)